Merge following kepler ds changes into the operator:

- sustainable-computing-io/kepler#578 - sustainable-computing-io/kepler#683 - sustainable-computing-io/kepler#635 - sustainable-computing-io/kepler#733 Signed-off-by: Parul Singh <[email protected]>
husky-parul · Jun 30, 2023 · a16d019 · a16d019
1 parent 0d76d4d
commit a16d019
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 6 deletions.
diff --git a/config/manager/kustomization.yaml b/config/manager/kustomization.yaml
@@ -5,4 +5,4 @@ kind: Kustomization
 images:
 - name: controller
   newName: quay.io/sustainable_computing_io/kepler-operator
-  newTag: 0.4.1
+  newTag: "0.5"
diff --git a/config/samples/kepler.system_v1alpha1_kepler.yaml b/config/samples/kepler.system_v1alpha1_kepler.yaml
@@ -13,8 +13,6 @@ spec:
   collector:
     image: quay.io/sustainable_computing_io/kepler:latest
     collectorPort: 9103
-  model-server-exporter:
-    port: 8100
 
 
 

diff --git a/controllers/kepler_controller.go b/controllers/kepler_controller.go
@@ -309,7 +309,7 @@ func (r *collectorReconciler) ensureConfigMap(l klog.Logger) (bool, error) {
 		var data_map = make(map[string]string)
 
 		data_map["KEPLER_NAMESPACE"] = r.Instance.Namespace
-		data_map["KEPLER_LOG_LEVEL"] = "5"
+		data_map["KEPLER_LOG_LEVEL"] = "1"
 		data_map["METRIC_PATH"] = "/metrics"
 		data_map["BIND_ADDRESS"] = bindAddress
 		data_map["ENABLE_GPU"] = "true"
@@ -318,6 +318,8 @@ func (r *collectorReconciler) ensureConfigMap(l klog.Logger) (bool, error) {
 		data_map["CGROUP_METRICS"] = "*"
 		data_map["MODEL_CONFIG"] = "| CONTAINER_COMPONENTS_ESTIMATOR=false CONTAINER_COMPONENTS_INIT_URL=https://raw.githubusercontent.com/sustainable-computing-io/kepler-model-server/main/tests/test_models/DynComponentModelWeight/CgroupOnly/ScikitMixed/ScikitMixed.json"
 
+		data_map["EXPOSE_HW_COUNTER_METRICS"] = "true"
+		data_map["EXPOSE_CGROUP_METRICS"] = "true"
 		r.configMap.Data = data_map
 
 		return nil
@@ -380,7 +382,7 @@ func (r *collectorReconciler) ensureDaemonSet(l klog.Logger) (bool, error) {
 				Privileged: &scc_value,
 			},
 			Image:   image,
-			Command: []string{"/usr/bin/kepler", "-address", bindAddress, "-enable-gpu=true", "enable-cgroup-id=true", "v=5"},
+			Command: []string{"/usr/bin/kepler", "-address", bindAddress, "-enable-gpu=true", "enable-cgroup-id=true", "v=5", "-kernel-source-dir=/usr/share/kepler/kernel_sources"},
 			Ports: []corev1.ContainerPort{{
 				ContainerPort: collectorPort,
 				Name:          "http",
@@ -411,8 +413,14 @@ func (r *collectorReconciler) ensureDaemonSet(l klog.Logger) (bool, error) {
 				FieldPath: "status.hostIP",
 			},
 		}
+		envFromSourceNode := corev1.EnvVarSource{
+			FieldRef: &corev1.ObjectFieldSelector{
+				FieldPath: "spec.nodeName",
+			},
+		}
 		r.daemonSet.Spec.Template.Spec.Containers[0].Env = []corev1.EnvVar{
 			{Name: "NODE_IP", ValueFrom: &envFromSource},
+			{Name: "NODE_NAME", ValueFrom: &envFromSourceNode},
 		}
 
 		r.daemonSet.Spec.Template.Spec.Containers[0].VolumeMounts = []corev1.VolumeMount{

diff --git a/controllers/sahandler.go b/controllers/sahandler.go
@@ -130,7 +130,7 @@ func (d *keplerSADescription) createOrUpdateClusterRole(l klog.Logger) (*rbacv1.
 		Rules: []rbacv1.PolicyRule{
 			{
 				APIGroups: []string{""},
-				Resources: []string{"nodes/metrics", "nodes/proxy", "nodes/stats"},
+				Resources: []string{"nodes/metrics", "nodes/proxy", "nodes/stats", "pods"},
 				Verbs:     []string{"get", "watch", "list"},
 			},
 		},